The Preliminary Applicability Evaluation Of PI-RADS V2 Diagnostic Score In 3.0T Mp-MRI Combined With PSAD For Prostate Cancer

Objective: To preliminarily evaluate the clinical value of Prostate Imaging Reporting and Data System version 2(PI-RADS v2)diagnostic score based on 3.0T multi-parameter magnetic resonance imaging(Mp-MRI)combined with prostate specific antigen density(PSAD)for prostate cancer(PCa)and preliminarily discuss the correlation between the predictive model established by PI-RADS v2 score combined with PSAD and the results of the new prostate cancer Grading Groups.Materials and methods: Collect the MRI and clinical data(including the patients' age and tPSA value)of 247 patients of clinical suspected prostate disease who meet the inclusion criteria and were confirmed pathologically by transrectal ultrasound(TRUS)guided prostate systematic puncture biopsies.All patients were underwent 3.0T Mp-MRI from June 2015 to November 2016.All the MRI sequences conclude high-resolution axial T2 WI combined with two kinds of functional imaging sequences(DWI and DCE-MRI),with phased-array body coil,and not using the endorectal coil.All MRI examinations were performed before biopsy or surgery,freezing,radiation,chemotherapy and endocrine treatment.According to the PI-RADS v2 proposed in 2014 Radiological Society of North America(RSNA),two MRI radiologists who were blinded to the clinical informations and pathologic results of the patients scored for the prostate MRI images of the 247 patients,respectively.The inconsistent scoring results were discussed again by the two radiologists to reach aggreement and then formed the final score.The final score of multifocal cases was determined by the score of dominant lesion(the highest PI-RADS score or with extracapsular extension or with the largest diameter).At the same time,measure the maximal anterior-posterior diameter and maximal left-right diameter of the prostate in the axial T2 WI and measure the maximal upper-lower diameter in the sagittal T2 WI,and then calculate the volume of prostate according to the calculative formula for prostatic volume(long ellipsoid formula): Volume(ml)=maximal anterior-posterior diameter(cm)× maximal left-right diameter(cm)× maximal upper-lower diameter(cm)×0.52,and the PSAD value according to the formula(PSAD=tPSA/V).According to the pathological diagnosis and Gleason score results,all cases which meet the inclusion criteria were divided into two groups: 1)the prostate cancer group and non cancer group;2)middle/high grade prostate cancer group(PCa with Gleason score ?7)and non middle/high grade prostate cancer group(including non cancer and PCa with Gleason score of <7).Statistical analysis was performed on all observed indicators of all included cases,including age,prostate volume,tPSA value,PSAD value,and PI-RADS v2 diagnostic score of the 247 patients.The statistical analysis was performed using SPSS 17.0 and Medcalc 15.0 statistical software,using two-sided test.Firstly,all observed indicators were judged whether or not in accordance with normal distribution using one sample K-S test.When P>0.05,it was considered in accordance with normal distribution.The univariate analysis was perfomed using independent sample t test,which was used in the variates in accordance with normal distribution and Wilcoxon two sample test,which was used in the variates not in accordance with normal distribution to compare the statistic difference of each variate between PCa and non cancer group and between middle/high grade prostate cancer group and non middle/high grade prostate cancer group.When P<0.05,the difference was considered statistically significant.Secondly,the receiver operating characteristic curve(ROC curve)of the variates with statistic difference in the univariate analysis were drew respectively and the area under ROC curve(AUC)values were calculated at the same time.The multivariate analysis of observed indicators with statistic difference in the univariate analysis was perfomed using the multivariate Logistic stepwise regression analysis to determine the independent predictors for PCa and middle/high grade prostate cancer.When P<0.05,it was considerde the independent predictor.Then,the Logistic regression models were established by the independent predictors for combined diagnosis of PCa(model 1)and middle/high grade prostate cancer(model 2).The datas of the independent predictors were brought into the two models to get the combined predictive probability(P).The ROC curves of the two models(model 1 and model 2)to diagnose PCa and middle/high grade prostate cancer were drew respectively by the Logit(P)and the AUC values were calculated at the same time.Z test was used to compare the statistical difference of the AUC value between all the indicators with statistical difference between groups(PCa and non cancer group,middle/high grade prostate cancer and non middle/high grade prostate cancer group)and at the same time,compare the statistical difference of the AUC value between the independent predictors and the two combined predictive models for PCa and middle/high grade prostate cancer,respectively.When P<0.05,the difference was considered statistically significant.The best threshold to diagonse PCa and middle/high grade prostate cancer and the sensitivity,specificity,negative predictive value,positive predictive value and Yuden index on the best thresholds of the independent predictors and the two combined models to diagonse the PCa and middle/high grade prostate cancer,were determined respectively by ROC curve analysis.Finally,Spearman correlation analysis was used to assess the correlation between PSAD,PI-RADS V2 score,the two combined predictive models and the new prostate Grading Groups.When P<0.05,it showed significant correlation.If the Spearman's coefficient was positive,the correlation was considered positive;if the Spearman's coefficient was negative,the correlation was considered negative.Results: A total of 247 patients who meet the inclusion criteria were enrolled in this study: PCa 110 cases,including low-grade cancer(Gleason score <7)23 cases and middle/high-grade cancer(Gleason score ?7)87cases;non PCa 137 cases,including benign prostatic hyperplasia 95 cases,acute or chronic prostatitis 34 cases,low grade intraepithelial neoplasia 4 cases and 4 normal cases.The results of one sample K-S test showed that the variate(age)was in accordance with normal distribution(P>0.05)and the other variates were not in accordance with normal distribution(P<0.05).Then,the univariate analysis was performed,using independent samples t test for age and Wilcoxon two sample test for the prostate volume,tPSA value,PSAD value and PI-RADS v2 scores.The univariate analysis showed that the differences of all these variates between PCa and non cancer group and between middle/high-grade cancer and non middle/high grade cancer group were all statistically significant(P<0.05).Multivariate analysis was performed using Logistic stepwise regression analysis for these variates with statistically significant differences in the univariate analysis and the results showed that PSAD value and PI-RADS V2 scores were independent predictors of PCa and middle/high grade prostate cancer(P<0.05).The two independent predictors were brought into multivariate Logistic forced regression analysis to establish the combined predictive models for PCa(model 1)and middle/high grade prostate cancer(model 2):model 1:Logit(P)=-5.097+2.309×PSAD+1.214×PI-RADS v2 soremodel 2:Logit(P)=-5.422+0.809×PSAD+1.219×PI-RADS v2 soreArea under the ROC curve(AUC)value of age,prostate volume,tPSA value,PSAD value and PI-RADS v2 score to diagnose PCa alone were: 0.61,0.657,0.788,0.851 and 0.886,respectively and the AUC value of the combined predictive model to diagnose PCa(model 1)was 0.911.The diagnostic performance of PI-RADS v2 score to diagnose PCa was highest in all observed indicators,but the difference of diagnostic performance between PI-RADS v2 score and PSAD value was not statistically significant(P>0.05).The diagnostic performance of PSAD value and PI-RADS v2 score to diagnose PCa were higher than age,prostate volume and tPSA value(P<0.05).The combined diagnostic performance of PI-RADS v2 score and PSAD value was higher than that of PI-RADS v2 score and PSAD value alone(P<0.05).The best thresholds of PSAD value,PI-RADS V2 score and the combined predictive model to diagnose PCa(model 1 Logit(P)value)were: 0.31ng/ml/ml,4 and-0.82 respectively;the sensitivity to diagnose PCa of PSAD value,PI-RADS V2 score and model 1 on the best threshold were: 0.764,0.782 and 0.855,respectively;the specificity were 0.854,0.912 and 0.847,repectively;the positive predictive value were 0.81,0.878 and 0.817,respectively;the negative predictive value were 0.824,0.839 and 0.879,respectively and the Yuden Index were 0.61,0.694 and 0.701,respectively.Area under ROC curve(AUC)value of age,prostate volume,tPSA value,PSAD value and PI-RADSv2 score to diagnose middle/high grade prostate cancer was: 0.599,0.652,0.812,0.883 and 0.886,respectively and the AUC value of the combined predictive model to diagnose middle/high grade prostate cancer(model 2)was 0.919.The diagnostic performance of PI-RADS v2 score to diagnose middle/high grade prostate cancer was highest in all observed indicators and the difference of diagnostic performance between PI-RADS v2 score and PSAD value was not statistically significant(P>0.05).The diagnostic performance of PI-RADS v2 score and PSAD value to diagnose middle/high grade prostate cancer alone were higher than that of age,prostate volume and tPSA value(P<0.05).The diagnostic performance of combined predictive model(model 2)was higher than that of PI-RADS v2 score alone(P<0.05),and there was no statistically significant difference of the diagnostic performance between model 2 and PSAD value to diagnose middle/high grade prostate cancer(P>0.05).The best threshold of PSAD value,PI-RADS v2 score and the model 2 to diagnose middle/high grade prostate cancer were 0.4ng/ml/ml,4 and-0.9,respectively;the sensitivity to diagnose middle/high grade prostate cancer of PSAD value,PI-RADS v2 score and model 2 on the best threshold were: 0.747,0.839 and 0.862,respectively;the specificity were 0.9,0.844 and 0.844,respectively;the positive predictive value were: 0.802,0.745 and 0.75,respectively;the negative predictive values were: 0.867,0.906 and 0.918,respectively and the Yuden Index were: 0.647,0.683 and 0.706,respectively.PSAD value,PI-RADS v2 score and the two combined predictive models(model 1 and model 2)all showed positive correlation on some degree with the new protate cancer Grading Groups results(Spearman's coefficient were 0.406,0.362,0.432 and 0.424,respectively,P<0.001);and the Spearman's coefficient of the two combined predictive models were higher than PSAD value and PI-RADS v2 score.Conclusion: 3.0T Mp-MRI PI-RADS v2 score and PSAD value were the independent predictive indicators for PCa and middle/high grade prostate cancer.The diagnositic performance of PI-RADS v2 score and PSAD value for PCa and middle/high grade prostate cancer were significantly higher than other clinical indicators(age,prostate volume and tPSA value).The diagnositic performance of PI-RADS v2 score combined with PSAD value for PCa was superior to that of PI-RADS v2 score and PSAD value alone.PI-RADS v2 score combined with PSAD value significantly improved the diagnostic performance of PI-RADS v2 score for middle/high grade prostate cancer alone.Combined application of PI-RADS v2 score and PSAD value can help to improve the diagostic sensitivity and the negative predictive value of PI-RADS v2 score and PSAD value alone and maintain the diagnostic specificity for PCa and middle/high grade prostate cancer.PSAD value,PI-RADS V2 score and the two combined predictive models(model 1 and model 2)all showed positive correlation on some degree with the new protate cancer Grading Groups results.The correlation between the two models(model 1 and model 2)and the new protate cancer Grading Groups was higher than PI-RADS v2 score and PSAD value.Combined application of PI-RADS v2 score and PSAD value can help predicting the invasiveness of PCa and guiding clinical decisions.